GitHub
Connect AI assistants to GitHub - manage repos, issues, PRs, and workflows through natural language.
作者提供的说明(英文原文)
Azure + GCP cost management: spending, forecasts, anomalies, budgets, idle resources, and tags.
这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。
我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。
选择你使用的 AI 应用,然后复制设置。
AZURE_SUBSCRIPTION_ID, AZURE_TENANT_ID, AZURE_CLIENT_ID, AZURE_CLIENT_SECRET, GOOGLE_CLOUD_PROJECT, GCP_BILLING_TABLE, GCP_PROJECT_ID, GCP_BILLING_ACCOUNT_ID, GOOGLE_APPLICATION_CREDENTIALS标为机密的值是你账户的钥匙。千万不要分享给别人,也不要粘贴到聊天里。
你需要
你用的是哪个应用?
在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。
打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json
{
"mcpServers": {
"cloudscope": {
"command": "npx",
"args": [
"-y",
"cloudscope-mcp"
],
"env": {
"AZURE_SUBSCRIPTION_ID": "your-value",
"AZURE_TENANT_ID": "your-value",
"AZURE_CLIENT_ID": "your-value",
"AZURE_CLIENT_SECRET": "your-value",
"GOOGLE_CLOUD_PROJECT": "your-value",
"GCP_BILLING_TABLE": "your-value",
"GCP_PROJECT_ID": "your-value",
"GCP_BILLING_ACCOUNT_ID": "your-value",
"GOOGLE_APPLICATION_CREDENTIALS": "your-value"
}
}
}
}在终端中运行这条命令
claude mcp add cloudscope --env AZURE_SUBSCRIPTION_ID=your-value --env AZURE_TENANT_ID=your-value --env AZURE_CLIENT_ID=your-value --env AZURE_CLIENT_SECRET=your-value --env GOOGLE_CLOUD_PROJECT=your-value --env GCP_BILLING_TABLE=your-value --env GCP_PROJECT_ID=your-value --env GCP_BILLING_ACCOUNT_ID=your-value --env GOOGLE_APPLICATION_CREDENTIALS=your-value -- npx -y cloudscope-mcp运行前请先填写空着的值。
粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"cloudscope": {
"command": "npx",
"args": [
"-y",
"cloudscope-mcp"
],
"env": {
"AZURE_SUBSCRIPTION_ID": "your-value",
"AZURE_TENANT_ID": "your-value",
"AZURE_CLIENT_ID": "your-value",
"AZURE_CLIENT_SECRET": "your-value",
"GOOGLE_CLOUD_PROJECT": "your-value",
"GCP_BILLING_TABLE": "your-value",
"GCP_PROJECT_ID": "your-value",
"GCP_BILLING_ACCOUNT_ID": "your-value",
"GOOGLE_APPLICATION_CREDENTIALS": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"cloudscope": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"cloudscope-mcp"
],
"env": {
"AZURE_SUBSCRIPTION_ID": "your-value",
"AZURE_TENANT_ID": "your-value",
"AZURE_CLIENT_ID": "your-value",
"AZURE_CLIENT_SECRET": "your-value",
"GOOGLE_CLOUD_PROJECT": "your-value",
"GCP_BILLING_TABLE": "your-value",
"GCP_PROJECT_ID": "your-value",
"GCP_BILLING_ACCOUNT_ID": "your-value",
"GOOGLE_APPLICATION_CREDENTIALS": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
这个服务器在你的电脑上运行,而 ChatGPT 只能连接互联网上的服务器。请选择其他应用。
你的选择通过 Cookie 保存在这台设备上,无需账户。
AZURE_SUBSCRIPTION_ID envAZURE_TENANT_ID envAZURE_CLIENT_ID envGOOGLE_CLOUD_PROJECT envGCP_BILLING_TABLE envGCP_PROJECT_ID envGCP_BILLING_ACCOUNT_ID env这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
它在你的电脑上运行;安装后,工具列表会显示在你的 AI 应用中。
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月6日
注册名称已通过 GitHub 账户 alexpota 验证,该账户拥有此仓库
我们还没有从 GitHub 读取这个仓库
我们还没有从 GitHub 读取这个仓库
它在你的电脑上运行,没有互联网端点
在你的电脑上运行,改用本地设置
它在你的电脑上运行;我们只能读取互联网端点的工具列表
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
它在 1 项适用的基础检查中通过了 1 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
你需要填写:AZURE_CLIENT_SECRET, GOOGLE_APPLICATION_CREDENTIALS
它在你的电脑上运行。你的 AI 应用会在需要时启动它,数据不会经过作者的服务器(除非它自己调用在线服务)。
GitHub 仓库最后更新于–,最新版本(0.3.3)于5个月前进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。你的电脑上还需要有 Node.js。
同一分类中有:GitHub, Desktop Commander, Cloudflare MCP
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
Connect AI assistants to GitHub - manage repos, issues, PRs, and workflows through natural language.
MCP server for terminal commands, file operations, and process management
Cloudflare MCP servers
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
Ask your AI about your cloud bill.
Ask once: "Give me this month's cost review" — and CloudScope runs the full workflow (spend, deltas, anomalies, idle resources, budgets, forecast) and hands back a decision-ready report, not raw numbers to interpret yourself:
## TL;DR
Month-to-date Azure spend is $4,312 (May 1–29), tracking +11% vs April.
The mover is App Service (+$280) — a new staging slot left running.
2 idle resources cost $94/month;
prod-budgetis at 78% with 2 days left.
Top action: delete the orphaned managed disk → save $63/month.
## Spend This Month
| Service | Cost | % |
|------------------|----------|------|
| Virtual Machines | $2,140 | 49.6% |
| App Service | $890 | 20.6% |
| Storage | $612 | 14.2% |
| Redis Cache | $410 | 9.5% |
| Networking | $260 | 6.0% |
## vs Last Month
+$427 (+11%) overall. App Service +$280 drove most of it; everything
else flat. → *new
staging-slot-2, created May 12, never torn down.*
## Idle & Wasted Resources
| Resource | Type | $/month |
|-----------------------|-------------|---------|
|
orphaned-osdisk-01| Managed Disk| $63 |
|
old-lb-ip| Public IP | $31 |
## Budget Status
prod-budget— 78% used, $4,312 / $5,500, on track (2 days left).
## Recommended Actions
1. Delete
orphaned-osdisk-01(unattached 40 days) → $63/month
2. Tear down
staging-slot-2if staging is idle → ~$280/month
3. Release
old-lb-ip(no binding) → $31/month
*Illustrative example showing the report format. Run /cloudscope:monthly-cost-review against your own subscription for live numbers.*
Azure is auto-detected from your
az loginsession. GCP requires BigQuery billing export setup (see below).
| Provider | Status |
|---|---|
| Azure | ✅ Supported |
| GCP | ✅ Supported |
| AWS | Use AWS's official server |
az login)gcloud auth application-default login)CloudScope auto-detects your subscription from az login. Just add the server:
{
"mcpServers": { "cloudscope": { "command": "npx", "args": ["-y", "cloudscope-mcp"] } } }
To target a specific subscription, add an env block:
{
"mcpServers": { "cloudscope": { "command": "npx", "args": ["-y", "cloudscope-mcp"], "env": { "AZURE_SUBSCRIPTION_ID": "your-subscription-id" } } } }
Advanced: Service Principal (CI/CD & automated environments)
| Variable | Description |
|---|---|
AZURE_TENANT_ID | Azure AD tenant ID |
AZURE_CLIENT_ID | App registration client ID |
AZURE_CLIENT_SECRET | App registration client secret |
Set these alongside AZURE_SUBSCRIPTION_ID in the env block above.
GCP requires a BigQuery billing export table. Find your table name in GCP Console > Billing > Billing export > BigQuery export.
{
"mcpServers": { "cloudscope": { "command": "npx", "args": ["-y", "cloudscope-mcp"], "env": { "GOOGLE_CLOUD_PROJECT": "my-project", "GCP_BILLING_TABLE": "my-project.my_dataset.gcp_billing_export_resource_v1_XXXXXX" } } } }
Set GOOGLE_CLOUD_PROJECT to your GCP project ID. Override with GCP_PROJECT_ID if your billing dataset lives in a different project.
These variables apply only when using GCP; Azure-only users can skip them.
| Variable | Description | Required |
|---|---|---|
GOOGLE_CLOUD_PROJECT | GCP project ID | Yes |
GCP_BILLING_TABLE | Fully-qualified BigQuery table (project.dataset.table) | Yes |
GCP_PROJECT_ID | Override project ID if different from GOOGLE_CLOUD_PROJECT | No |
GCP_BILLING_ACCOUNT_ID | Billing account ID for budget monitoring | No |
GOOGLE_APPLICATION_CREDENTIALS | Path to service account JSON key file | No |
{
"mcpServers": { "cloudscope": { "command": "npx", "args": ["-y", "cloudscope-mcp"], "env": { "AZURE_SUBSCRIPTION_ID": "your-subscription-id", "GOOGLE_CLOUD_PROJECT": "my-project", "GCP_BILLING_TABLE": "my-project.my_dataset.gcp_billing_export_resource_v1_XXXXXX" } } } }
Azure is auto-detected from az login. Add AZURE_SUBSCRIPTION_ID to target a specific subscription.
CloudScope is read-only. It cannot create, modify, or delete any cloud resources. Azure uses Cost Management Reader permissions. GCP uses BigQuery Data Viewer + Job User with no write access.
Does this modify my cloud resources? No. Read-only access only.
Do I need a service principal? No. az login (Azure) or gcloud auth application-default login (GCP) works for local use.
Does the Azure Cost Management API cost money? No. It's free.
Do GCP BigQuery cost queries cost money? Yes, but typically <$0.01 per query ($6.25/TiB scanned, first 1 TiB/month free). See BigQuery pricing.
Can I use both Azure and GCP at the same time? Yes. Configure both sets of env vars and CloudScope queries whichever provider you specify in each tool call.
git clone https://github.com/alexpota/cloudscope-mcp.git
cd cloudscope-mcp npm install npm run build npm test npx @modelcontextprotocol/inspector node dist/index.js
Most tools accept a provider parameter (azure or gcp); the provider-specific tools (list_subscriptions, list_projects, and the cross-account tools get_cross_subscription_costs / get_cross_project_costs) are fixed to their own cloud. The default provider is auto-detected based on which providers are configured.
Cost Analysis
| Tool | Description | Key Parameters |
|---|---|---|
get_cost_summary | Spending breakdown by service, group, or region | start_date, end_date, group_by |
get_cost_by_tag | Costs grouped by a tag/label key | tag_key, start_date, end_date |
compare_periods | Side-by-side cost comparison of two date ranges | period_a_start/end, period_b_start/end |
top_spending_resources | Most expensive individual resources | days, limit |
get_cross_subscription_costs | Combined costs across Azure subscriptions | subscription_ids, start_date, end_date |
get_cross_project_costs | Combined costs across GCP projects | project_ids, start_date, end_date |
Monitoring
| Tool | Description | Key Parameters |
|---|---|---|
detect_anomalies | Find spending spikes vs previous period | days, threshold |
check_budgets | Budget status, current spend, projected overage | _(none)_ |
get_cost_forecast | Predict spending based on current trends | days |
Optimization
| Tool | Description | Key Parameters |
|---|---|---|
list_recommendations | Cost optimization suggestions (Azure Advisor / GCP Recommender) | category |
find_idle_resources | Provisioned but unused resources with cost estimates | _(none)_ |
find_untagged_resources | Resources with no tags/labels (cost attribution gaps) | _(none)_ |
Utility
| Tool | Description | Key Parameters |
|---|
| get_current_date | Today's date and current/previous month
只显示了开头部分;其余内容请在仓库中阅读